AGENT SKILL

audit-reporting

Structured finding format: severity, impact, PoC, remediation. Templates and classification standards for audit output. Used by Report Compiler.

License 许可
MIT
最近更新
2026-10-18
Stars
★ 0
Forks
0

01 前言 Front Matter

name
audit-reporting
description
Structured finding format: severity, impact, PoC, remediation. Templates and classification standards for audit output. Used by Report Compiler.

03 SKILL.md 正文

Audit Reporting Skill

Purpose

Ensure all audit outputs follow a consistent, auditor-grade format with proper severity classification, concrete impact descriptions, and traceable evidence chains.

Severity Classification (Immunefi Standard)

Severity Criteria
Critical Direct loss of funds, permanent freezing of funds, protocol insolvency
High Theft of unclaimed yield, permanent DoS, governance manipulation
Medium Griefing with material loss, temporary DoS, value extraction below threshold
Low Non-critical issues, informational findings, gas optimizations

Finding Report Structure

Every confirmed finding must include:

  1. Title: One-line description
  2. Severity: Critical/High/Medium/Low with justification
  3. Impact: Concrete description (e.g., "Attacker can drain 100% of vault TVL")
  4. Root Cause: Specific code flaw with file and line reference
  5. Attack Path: Step-by-step, numbered sequence
  6. PoC: Link to runnable Foundry test
  7. Affected Code: File, lines, snippet
  8. Remediation: Specific fix with code example

Evidence Chain

Every finding must have a traceable path:

Task (specs/tasks/TASK-{N}.md)
  → Reasoning (specs/reasoning/TASK-{N}-reasoning.md)
    → Verification (specs/verification/TASK-{N}-verification.md)
      → Finding (findings/FINDING-{ID}.md)

What Makes a Finding Actionable

  • ❌ "There may be a reentrancy risk" — NOT actionable
  • ✅ "Reentrancy in Vault.withdraw() allows attacker to drain all ETH. PoC at test/exploit.t.sol" — actionable

Coverage Reporting

The coverage map must honestly distinguish:

  • Full coverage: Function analyzed with multiple task approaches
  • Partial coverage: Function analyzed but not all paths/values tested
  • No coverage: Function not targeted by any task (must be in gap record)

04 同仓库更多 Skills